基于时延特性的网络拓扑推断技术研究

发布时间：2018-04-10 02:19

本文选题：网络断层扫描　切入点：深度优先搜索　出处：《兰州交通大学》2014年硕士论文

【摘要】：现如今网络规模和网络复杂性日益增长，在互联网的研究中，准确和及时的识别路由器级互联网的拓扑结构是目前研究的热点和难点。运用网络断层扫描技术来推断网络拓扑结构，是一种极具吸引力的方法。网络断层扫描(Network Tomography,NT)是一种采用全新思想的网络测量技术,它是基于医学透视的概念提出来的。该技术是一种基于端到端的测量方法，通过反演推断获取网络的内部特性，并运用统计和推断的原理进行拓扑结构的推测。网络断层扫描技术不需要网络内部的节点的合作，只需要选择一组接收节点即可完成对网络内部特性的推断，解决了网络内部节点不协作的问题。然而，目前断层扫描技术有一定的限制性，不能及时推断出大规模的网络拓扑结构，因为推断大规模网络拓扑结构需要的测量工程较大，难以实现。本文对基于时延特性的网络拓扑推断技术进行研究，描述了一种基于叶节点深度优先搜索(DFS)序列的网络拓扑推断算法，首先使用递归二分法找出叶节点DFS序列，然后运用叶节点FS序列推断网络拓扑结构。该算法能够高效的发现网络的逻辑拓扑树结构。通过大量的仿真实验表明，该方法所需要的成对探测包的数量低于传统聚类算法的15%，可以更准确、更快速的推断出网络的拓扑结构，进而更大程度的提高了网络拓扑推断的效率。论文具体工作如下：前三章介绍了课题的研究背景、研究现状以及研究内容，并介绍了本课题研究中用到的网络测量技术和网络断层扫描技术的相关内容，并且对现有的较成熟的网络拓扑推断算法进行了概括和总结。第四章为本文核心内容，主要介绍了基于叶节点DFS序列的网络拓扑推断算法，，并且详细描述了使用该算法推断网络拓扑结构的整个过程。第五章仿真实验，使用NS2仿真平台，对第四章提出的基于叶节点DFS序列的网络拓扑推断算法进行了有效验证，证明了该算法的高效性和可行性。
[Abstract]:Nowadays, network scale and network complexity are increasing day by day. In the research of Internet, accurate and timely identification of router level Internet topology is a hot and difficult point.It is an attractive method to infer the network topology by using the network tomographic technique.Network Tomography (NT) is a new network measurement technology based on the concept of medical perspective.This technique is an end-to-end measurement method, which obtains the internal characteristics of the network by inversion inference, and uses the principles of statistics and inference to infer the topology of the network.The network tomography technology does not need the cooperation of the nodes in the network, but only needs to select a group of receiving nodes to complete the inference of the internal characteristics of the network, which solves the problem of the nodes in the network not cooperating.However, at present, the technique of tomographic scanning is too restrictive to infer the large-scale network topology in time, because it is difficult to realize the large scale network topology because the measurement engineering is needed to infer the large-scale network topology structure.In this paper, the network topology inference technology based on delay characteristic is studied, and a network topology inference algorithm based on the leaf node depth first search sequence is described. Firstly, the DFS sequence of the leaf node is found by recursive dichotomy.Then the topological structure of the network is inferred by using the leaf node FS sequence.The algorithm can efficiently discover the logical topology tree of the network.A large number of simulation experiments show that the number of pairwise detection packets in this method is lower than that of the traditional clustering algorithm, and the topology of the network can be inferred more accurately and quickly.Furthermore, the efficiency of network topology inference is improved to a greater extent.The specific work of the thesis is as follows:The first three chapters introduce the research background, research status and research content, and introduce the network measurement technology and network tomography technology used in this research.The existing network topology inference algorithms are summarized and summarized.The fourth chapter is the core of this paper, mainly introduces the network topology inference algorithm based on the leaf node DFS sequence, and describes the whole process of using the algorithm to infer the network topology structure in detail.In the fifth chapter, using the NS2 simulation platform, the network topology inference algorithm based on the leaf node DFS sequence is validated, and the efficiency and feasibility of the algorithm are proved.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】